View
3
Download
0
Category
Preview:
Citation preview
East Tennessee State UniversityDigital Commons @ East
Tennessee State University
Electronic Theses and Dissertations Student Works
5-2017
Black Bears (Ursus americanus) versus BrownBears (U. arctos): Combining Morphometrics andNiche Modeling to Differentiate Species andPredict Distributions Through TimeTheron Michael KantelisEast Tennessee State University
Follow this and additional works at: https://dc.etsu.edu/etd
Part of the Other Environmental Sciences Commons, Paleobiology Commons, and thePaleontology Commons
This Thesis - Open Access is brought to you for free and open access by the Student Works at Digital Commons @ East Tennessee State University. Ithas been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of Digital Commons @ East Tennessee StateUniversity. For more information, please contact digilib@etsu.edu.
Recommended CitationKantelis, Theron Michael, "Black Bears (Ursus americanus) versus Brown Bears (U. arctos): Combining Morphometrics and NicheModeling to Differentiate Species and Predict Distributions Through Time" (2017). Electronic Theses and Dissertations. Paper 3262.https://dc.etsu.edu/etd/3262
Black Bears (Ursus americanus) versus Brown Bears (U. arctos): Combining
Morphometrics and Niche Modeling to Differentiate Species and Predict Distributions
Through Time
A thesis
presented to
the faculty of the Department of Geosciences
East Tennessee State University
In partial fulfillment
of the requirements for the degree
Master of Science in Geosciences
_____________________
by
Theron Michael Kantelis
May 2017
_____________________
Dr. Blaine W. Schubert, Chair
Dr. Joshua Samuels
Dr. Chris Widga
Keywords: Ursus arctos, Ursus americanus, Bears, Geometric Morphometrics,
Ecological Niche Model, Teeth, Molars
2
ABSTRACT
Black Bears (Ursus americanus) versus Brown Bears (U. arctos): Combining
Morphometrics and Niche Modeling to Differentiate Species and Predict Distributions
Through Time
by
Theron Michael Kantelis
Late Pleistocene American black bears (Ursus americanus) often overlap in size with
Pleistocene brown bears (U. arctos), occasionally making them difficult to diagnose.
Large U. americanus have previously been distinguished from U. arctos by the length of
the upper second molar (M2). However, the teeth of fossil U. americanus sometimes
overlap size with U. arctos. As such, there is need for a more accurate tool to
distinguish the two species. Here, 2D geometric morphometrics is applied to the
occlusal surface of the M2 to further assess the utility of this tooth for distinguishing U.
americanus and U. arctos specimens. When combined with an Ecological Niche Model
of U. americanus and U. arctos in North America from the Last Glacial Maximum, this
morphometric technique can be applied to key regions. A case of two Pleistocene
specimens previously identified as U. arctos from eastern North America exemplifies the
utility of this combination.
3
TABLE OF CONTENTS
Page
ABSTRACT ..................................................................................................................... 2
LIST OF TABLES ............................................................................................................ 4
LIST OF FIGURES .......................................................................................................... 5
Chapter
1. INTRODUCTION ......................................................................................................... 6
2. GEOMETRIC MORPHOMETRIC ANALYSIS OF THE M2 OF URSUS
AMERICANUS AND U. ARCTOS ................................................................................. 12
Materials and Methods ....................................................................................... 12
Data Acquisition ....................................................................................... 12
Data Processing ...................................................................................... 17
Results ............................................................................................................... 18
Principal Component Analysis ................................................................. 18
Discriminant Functions............................................................................. 21
Thin Plate Spline and Averaged Images .................................................. 29
Discussion .......................................................................................................... 33
Morphological Diagnosis .......................................................................... 33
Further Implications ................................................................................. 34
3. ECOLOGICAL NICHE MODEL OF U. AMERICANUS AND U. ARCTOS
PROJECTED TO THE LAST GLACIAL MAXIMUM ...................................................... 37
Materials and Methods ....................................................................................... 38
Results ............................................................................................................... 40
Discussion .......................................................................................................... 50
4. CONCLUSIONS AND CASE STUDY ........................................................................ 52
BIBLIOGRAPHY ........................................................................................................... 63
APPENDIX .................................................................................................................... 68
Additional Table and Figures .............................................................................. 68
4
VITA .............................................................................................................................. 80
5
LIST OF TABLES
Table Page
1. Landmark Descriptions and Placement ..................................................................... 14
2. Total Variance Explained for Principal Component Analysis ..................................... 19
3. Component Matrix for the Principal Component Analysis ......................................... 22
4. Eigenvalues, Wilk’s Lambda, and Classification Results ........................................... 24
5. Stepwise Eigenvalues, Wilk’s Lambda, and Classification Results ........................... 27
6. Total Variance Explained for the Case Study
Principal Component Analysis ............................................................................ 55
7. Component Matrix for the Case Study Principal Component Analysis ...................... 58
8. Eigenvalues, Wilk’s Lambda, and Classification Results
of the Case Study ............................................................................................... 59
9. Stepwise Eigenvalues, Wilk’s Lambda, and Classification Results
of the Case Study ............................................................................................... 61
6
LIST OF FIGURES
Figure Page
1. Morphological Regions of the Upper Second Molar of Ursus americanus in
Occlusal View .................................................................................................... 10
2. Visual Display of Landmark Locations ...................................................................... 15
3. Scatter Plot of PC1 and PC2 .................................................................................... 20
4. Scatter Plot of PC1 and PC3 .................................................................................... 21
5. Histogram Plot of the Discriminant Function ............................................................ 25
6. Histogram Plot of the Stepwise Discriminant Function ............................................. 28
7. Thin Plate Spline ...................................................................................................... 30
8. Averaged Images of the M2 of U. americanus and U. arctos ................................... 31
9. Modern Ecological Niche Model for U. americanus .................................................. 41
10. Modern Ecological Niche Model for U. arctos ........................................................ 42
11. Last Glacial Maximum Ecological Niche Model for U. americanus ......................... 43
12. Last Glacial Maximum Ecological Niche Model for U. arctos ................................. 44
13. Sensitivity vs. Specificity charts for the Ecological Niche Models ........................... 45
14. Comparison of the U. americanus Modern and LGM
Ecological Niche Model ..................................................................................... 47
15. Comparison of the U. arctos Modern and LGM
Ecological Niche Model ..................................................................................... 48
16. Area of significant presence (>.5) overlap
between U. americanus and U. arctos at the LGM ............................................ 49
17. Scatter Plot of the Case Study PC1 and PC2 ........................................................ 56
18. Scatter Plot of the Case Study PC1 and PC3 ........................................................ 57
19. Histogram Plot of the Case Study Discriminant Function ....................................... 60
20. Histogram Plot of the Case Study Stepwise Discriminant Function........................ 62
7
CHAPTER 1
INTRODUCTION
Numerous Pleistocene fossil sites across North America have reported
specimens of Ursus americanus and Ursus arctos (Kurtén and Anderson, 1980;
Graham and Lundelius 2010). Occasionally, these two species of bear are even found
at the same fossil locality, and sometimes alongside other genera of bear such as
Arctodus or Tremarctos (Kurtén and Anderson, 1980; Graham and Lundelius 2010).
While neither U. americanus or U. arctos is easily mistaken for tremarctine bears, the
majority of the differences between U. americanus and U. arctos may be difficult to
assess in the fossil record (Gordon 1977, DeMaster and Stirling 1981, Graham 1991,
Pasitschniak-Arts 1993, Larivière 2001). Ursus arctos is described as having a dished
facial profile, significantly longer claws on the paws of the forelimbs than the hind limbs,
an upper second molar (M2) length greater than 31 mm, a lower first molar (m1) length
and width greater than 20.4 and 10.5 mm respectively, and a prominent shoulder hump,
whilst U. americanus possess no such hump, has claws of nearly equal length on all
paws, and has a more concave profile (DeMaster and Stirling 1981, Pasitschniak-Arts
1993, Larivière 2001). Aside from these features, modern U. arctos and U. americanus
can be distinguished from each other by the larger size of U. arctos and often simply by
the color of their pelage (DeMaster and Stirling 1981, Pasitschniak-Arts 1993, Larivière
2001). It should be noted, however, that while the common names of U. arctos and U.
americanus imply a simple brown and black coloration respectively, U. americanus has
been well documented to range widely in color (Pasitschniak-Arts 1993, Larivière 2001,
DeMaster and Stirling 1981).
8
While this is a fair list of distinguishing features for living members of U. arctos
and U. americanus, many of these characteristics are not necessarily applicable to
fossil specimens. Coloration is not typically available in paleontology as the pelage is
nearly always absent and is often discolored in the rare case of its preservation. The
facial profile is difficult to distinguish without fleshy features and particularly if the skull is
deformed or crushed. The shoulder hump present in U. arctos is not an osteological
feature, so it does not readily preserve. Size is not necessarily reliable as studies have
shown that Pleistocene and some Holocene U. americanus are able to achieve a size
comparable to both modern and fossil members of U. arctos (Kurtén and Anderson
1980, Wolverton and Lyman 1998). In addition to a general size increase, the size of the
teeth in Pleistocene U. americanus were also greater, making the use of molar length
and width in identification difficult (Graham 1991). This leaves the length of the front and
hind claws for the two species. In many cases, these elements are not discovered with
the fossil specimen, so an identification must be made based on what is available
(Elftman 1931, Kurtén 1963, Mustoe and Carlstad 1995, Czaplewski et al. 1999,
Czaplewski and Willsey 2013, Czaplewski and Puckette 2014). These identifications are
not made off of a defining character, but morphological similarity. As stated in several of
these articles, this strategy is not preferable due to this technique leaving the
identifications somewhat ambiguous.
As a result, there is a need for a greater number of diagnostic tools to identify
these species in the fossil record. Preferably, this would be a method which does not
rely on general morphological similarities. Due to the extreme level of general variability
found in bears, it may require a larger number of specimens to accurately represent the
9
typical morphology of the species (Baryshnikov 2006). While the statement of extreme
variability may sound extravagant, members of U. arctos alone have been attributed to
upwards of 80 different species as well as several different genera (Pasitschniak-Arts
1993). While all of these have been synonymized with U. arctos or now refer to
subspecies, this exemplifies how the variability within Ursus can cause confusion in
identification. Also, there is a precedent set for misidentified specimens. In 1991,
Graham suggested the reassignment of several Pleistocene Ursus due to
misunderstandings of how large U. americanus could be during that time period. These
specimens had been assigned to U. arctos based on their size, as no modern U.
americanus have reached that size. Despite the variation in Ursus, length and width
measurements of the molars has shown to be one of the most effective ways to identify
a specimen, fossil or otherwise (Graham 1991). Therefore, if a method is required to
better identify specimens, it logically would originate at the molars.
Gordon (1977) showed that modern U. arctos and U. americanus could be
distinguished by the length and width of m1 and M2. The length and width of the m1
had a 100 percent success rate in identifying the species and the length of the M2 had
95 and 100 percent success rate for the two species, respectively. Gordon used
between 51 and 144 specimens for each of these measurements, but neglected to
mention where these specimens were sourced from. While Gordon tested other
measurements of the teeth, these were either significantly less successful in one or both
species, or were shown to be inconsistent by Graham in 1991. Between the m1 and M2,
the M2 is the more common of the teeth due to its size and connection to skull rather
than the mandible. As such, the M2 is the tooth of choice in this study. Length and width
10
measurements of the teeth have shown to be somewhat ineffective due to an overlap in
size between Pleistocene U. americanus and U. arctos (Kurtén and Anderson 1980). As
such, a geometric morphometric study would be preferred to a study of linear
measurements; though there are complications. As previously stated, bears generally
have a high degree of variability. Within U. arctos and U. americanus, the number of
premolars is variable, and even between the left and right and maxillary and mandibular
premolars (Graham 1991, Baryshnikov et al. 2003, Baryshnikov 2006, Baryshnikov
2007). Gordon (1977) and Graham (1991) attempted to test the presence of accessory
cusps on the p4, m1, and P4, but were met with widely disparate success rates.
Dentral terminology here follows to Baryshnikov (2007) “for describing the ursine
M2. The different cusps and areas of a typical ursine M2 are detailed in Figure 1.
11
Figure 1. Morphological Regions of the Upper Second Molar of Ursus americanus in Occlusal View (ETVP 18252; left M2 shown). 1) Protocone, 2) Paracone, 3) Metaconule, 4) Metacone, 5) Hypocone, 6) Talon, 7) Post-Metacone Accessory Cusp (not always present), 8) Post-Hypocone Accessory Cusp (not always present). The shapes of these regions may be inconsistent between individuals and species, but their relative positions are consistent outside of pathologies.
Development of a technique for differentiating the M2 of U. arctos and U.
americanus is one aspect of this project. Ursus occurs throughout the majority of North
America, and fossil specimens from the Pleistocene appear to have been present within
a quite different range from that of modern times. (Pasitschniak-Arts 1993, Larivière
2001, Graham and Lundelius 2010). To better predict what species of Ursus would be
found at a particular locale, this study also creates an ecological niche model of where
12
U. arctos and U. americanus are expected to have been present during the Last Glacial
Maximum (LGM). By using modern bioclimatic variables to understand the habitat
preferences of these species, bioclimatic variables from the LGM can be used to
approximate the LGM range. This may help in identifying specimens whose
identification should be reassessed. Specimens occurring significantly outside of their
expected range could represent misidentification, or, if they were correctly identified, a
substantial difference between the environmental preferences of Ursus species
between the Pleistocene and now. There are periods of the Pleistocene which were
different ecologically from today and the LGM, so this model cannot be perfect, but this
model will give a general idea of where the species could have been present given the
general climatic differences between the Pleistocene and present. As there are fossil
specimens of that have been described as U. arctos that are well outside of their
modern historic range, this aspect of the study should help to determine how unusual or
expected that expanded range is (Graham and Lundelius 2010).
CHAPTER 2
13
GEOMETRIC MORPHOMETRIC ANALYSIS OF THE M2 OF URSUS AMERICANUS
AND U. ARCTOS
Materials and Methods
Data Acquisition
For the analysis, 64 (34 U. americanus, 30 U. arctos) modern specimens
(collected within the past 150 years) were collected from two collections, the East
Tennessee State University comparative collection (ETVP) and the National Museum of
Natural History’s (NMNH) mammal collection. From the ETVP, U. americanus
specimens are largely from Tennessee, and U. arctos are largely from Alaska, with
several belonging to U. a. middendorffi, the Kodiak subspecies. From the NMNH
collection, specimens originate from the entirety of the modern historic range. Specimen
numbers, and their origins can be seen in Appendix A (pg 64). The choice of specimens
for this study attempts to mimic and therefore account for any regional geographic
variation that might otherwise skew the results. This said, there aren’t any specimens in
the sample that represent regions outside North America. The inclusion of such
specimens could have given the results a “Eurasian” skew, and not accurately
represented the differences between bears that might have actually occurred in the
same region (Taberlet and Bouvet 1994).
The ETVP collection Ursus collection is limited in geographic scope, and most
available specimens were used that had complete M2’s with well-preserved cusps.
Because of the vast collection at the NMNH, specimen choice focused on those with
teeth in the best condition. Complete and relatively unworn were preferred, with careful
14
attention made to not disregard teeth that appeared to have an unusual shape or
morphology (not to be confused with those bearing a pathology). While such teeth might
be considered potential outliers in studies of most animals, bear teeth are well
documented to be highly variable (Baryshnikov 2007), so this line of thought could
potentially create a bias towards overly homologous teeth, when in reality, many teeth
of relatively strange or unusual shape may need to be identified by the results of this
study (Baryshnikov 2006).
Data acquisition consisted of photographs taken perpendicular to the palate,
such that the lens of the camera was parallel with the flat surface of the palate. Future
studies might use 3D landmarks (e.g., using a microscribe), but care will need to be
taken in landmark choice, as wear on the teeth could invalidate some of the landmarks
chosen here. That said, 3D landmarks remove the concern of photograph orientation,
so there are definite benefits to this strategy.
Table 1. Landmark Descriptions and Placement. See Figure 2 for a visual
representation of the placement of these landmarks.
15
Landmark Description
1 Point of maximum curvature on labial side of the paracone
2 Point of minimum curvature on labial side between the paracone and
metacone
3 Point of maximum curvature on labial side of the metacone
4 Point of maximum curvature on posterior end of tooth
5 Point of maximum curvature on lingual side of the anterior portion of
lingual cingulum
6 Most anterior tip of the paracone
7 Apex of paracone
8 Intersection of paracone and metacone blades
9 Apex of metacone
10 Most posterior end of the metacone
11 Point of maximum curvature on lingual side of the metacone
12 Intersection of the paracone, metacone, and protocone
13 Intersection of the protocone, metaconule, and metacone.
16
Figure 2. Visual Display of Landmark Locations. USNM 227660 (U. americanus)
displaying the locations of the thirteen landmarks used in this study.
17
As previously mentioned, landmarks were chosen based on consistency and
ability to be recognized, as such, all landmarks chosen are type 1 and type 2 (Bookstein
1991). Slinding landmarks and type 3 landmarks were avoided due to concerns of
consistency within each species (Bookstein 1991). The landmarks chosen are listed and
shown in Table 1, and Figure 2, respectively. Landmarks 1-6 were chosen to represent
the recognizable and consistent locations on the exterior outline of the tooth, these
being the widest point of the paracone, narrowest point between the paracone and
metacone, widest point at the metacone, the most posterior, most lingual, and most
anterior tips respectively. Landmarks 7 and 9 were chosen as they are the only cusp
apices which are consistently recognizable, as all lingual side cusp apices are difficult to
consistently recognize with a strong level of certainty. Only 35 of the 63 specimens
(53.84%) would have been able to have additional landmarks placed along the lingual
side cusps. The remaining specimens all had one or more potential landmark that could
not be confidently identified. The protocone long, ridgelike, and often split, obscuring or
duplicating the apex; the metaconule is reduced in U. americanus; the hypocone is
often reduced in U. americanus; and the post hypocone is often absent or highly
reduced in U. arctos. Landmark 8 and 10 were chosen to represent the boundaries of
the paracone and metacone alongside landmark 6. Landmarks 11-13 were chosen to
flesh out the remaining boundaries for the paracone and metacone, but also acted as
proxies for the size and location of the lingual cusps.
18
Data Processing
After photographs were taken, they were compiled into a TPS file using tpsUtil64
(Version 1.74), part of the “tps” series of programs developed by F. James Rohlf (2015).
This was done once for the teeth from ETSU, once for the teeth from NMNH, once to
encompass both of them. From here, photographs had landmarks placed on them using
tpsDIG2w32 software (Version 2.30) (Rohlf 2015). Any tooth whose landmarks could
not be confidently assessed during the landmarking process was removed from the
study. Once landmarked, the image sets had their consensus configuration determined
through tpsRelw32 (Version 1.67) (Rohlf 2015). The aligned landmark data was saved
from the consensus. After the aligned landmark data was saved, it was formatted using
Microsoft Office Excel such that it was in a format applicable in the SPSS statistical
package. Within SPSS, a trio of analyses were performed for each group of specimens:
A Principal Component Analysis (PCA), a Discriminant Analysis, and a Step-Wise
Discriminant Analysis. Deformation grids were added to the plots of the analyses for the
sake of representing the shape differences along the axes. These deformation grids
were generated by placing the principal component scores and discriminant function
scores in a .nts file for use within tpsRegr (Version 1.42) (Rohlf 2015).
In addition, a copy of the data file for specimens from the NMNH was split into U.
americanus and U. arctos files, from which unwarped and averaged images were
created in the tpsSuper (Version 2.03) and saved (Rohlf 2015). The averaged images
hold no real statistical significance, but they do help to provide a visual representation of
the differences between the two species. A thin-plate spline was also generated from
19
the NMNH specimens using the aligned U. americanus and U. arctos data from the
averaged images in tpsSpline (Version 1.22) (Rohlf 2015).
Results
Principal Component Analysis
The results from the principal component analysis consisted of eight principal
components scoring an eigen value of greater than 1, with the top three principal
components representing 50% of the cumulative variance at 27.867%, 12.851%, and
9.520%. As the first 3 principal components represent 50% of the variance total, and
proceeding components represent less than 10% of the variance each, scatter plots
were only generated for the first three principal components. As is described below,
principal components two and three appear to be showing within species variation, so
they were not plotted against each other. Data for these and the remaining principal
components can be seen in Table 2. A scatter plot of the first and second principal
components can be seen in Figure 3, and a plot of the first and third principal
components can be seen in Figure 4.
The first principal component appears to be expressing an elongation of the tooth
in U. arctos and an expansion of the lingual cusps, though the space between the
intersection of the paracone metacone and protocone and the intersection of the
protocone, metaconule, and metacone are is much smaller. An elongation of the tooth
in U. arctos is not surprising considering the usefulness of that measurement in modern
specimens (Gordon 1977, Graham 1991). The second principal component appears to
show variation of the relative position of the cusp apices within each species as it does
20
not show any strong variation between the species. The third principal component
appears to show variation of the length of the posterior portion of the metacone within
each species without any strong variation between species. The component matrix in
Table 3 shows the scores for these and the remaining components.
Table 2. Total Variance Explained for Principal Component Analysis
.
21
Figure 3. Scatter Plot of PC1 and PC2. The first principal component shows lateral
compression or extension with an inverse effect along the anterior landmarks. The
second principal component shows a vertical bending of the anterior landmarks of the
tooth.
22
Figure 4. Scatter Plot of PC1 and PC3. The first principal component shows lateral
compression or extension with an inverse effect along the anterior landmarks. The third
principal component shows a twisting such that the lingual and labial sides of the middle
of the tooth inversely compress or expand.
24
Discriminant Functions
The discriminant function analysis resulted in a high eigenvalue of 8.736 and a
.103 Wilk’s lambda. These indicate that large proportion of the variance is explained by
the discriminant function and that the two species have very low variability within their
groups compared to the variability between each other. Eigenvalues, Wilk’s lambda,
and Classification results can be seen in Table 4. The discriminant function showed a
94.1% success rate at identifying U. americanus, and a 93.3% success rate at
identifying U. arctos. Further inspection of the misidentified individuals suggests that
one of the misidentified specimens (ETVP 5162) may actually be identified within the
collection incorrectly, thus increasing the U. arctos identification success rate to 96.6%.
Images of the misidentified teeth, as well as the Casewise Statistics can be found in
Appendix A. The Histogram plot of the discriminant function can be seen in Figure 5.
The discriminant function appears to be representing the same shape variation as seen
in the first principal component
25
Table 4. Eigenvalues (Top), Wilk’s Lambda (Middle), and Classification Results
(Bottom). Classification results show a strong level of accuracy, and helped to reveal
one possibly incorrectly identified specimen.
26
Figure 5. Histogram Plot of the Discriminant Function. The discriminant function
represents the same variation as the first principal component.
27
Much like the discriminant function, the stepwise discriminant function analysis
resulted in a high, though lower than the discriminant function’s, eigenvalue of 5.059
and a low, though higher than the discriminant function’s, .103 Wilk’s lambda. The
stepwise discriminant specified variables Y3, Y10, Y13, Y8, and X6 as being particularly
valuable in discrimination. These variables, with the exception of X6 appear to be
clustered around the metacone and metaconule, suggesting that these cusps might be
important in distinguishing the teeth of the two species. Variables Entered, Eigenvalues,
Wilk’s lambda, and Classification results can be seen in Table 4. While the discriminant
function misidentified some specimens, the stepwise discriminant function had zero
misidentifications. This said, with ETVP 5162 potentially being misidentified within the
collection, the success rate of U. arctos identification remains at 96.66%. Casewise
Statistics can be found in Appendix A. The Histogram plot of the stepwise discriminant
function can be seen in Figure 6. The stepwise discriminant function appears to be
representing very similar shape variation as the discriminant function, though with more
expansion and contraction in the posterior half of the tooth.
28
Table 5. Stepwise Eigenvalues (Top), Wilk’s Lambda (Middle), and Classification
Results (Bottom). Classification results show a strong level of accuracy, though the
presence of a possibly misidentified specimen within the collection lowers the accuracy
of U. arctos identification to 96.66%.
29
Figure 6. Histogram Plot of the Stepwise Discriminant Function. The stepwise
discriminant function represents very similar variation as the discriminant function.
30
Thin Plate Spline and Averaged Images
The thin plate spline (Figure 7) shows several prominent differences between the
M2 of the two species. In relation to U. americanus, U. arctos does have a longer tooth,
as has been stated previously, but there are other prominent differences. This includes
the following characters for U. arctos that distinguish it from U. americanus: paracone
and metacone have much longer blades on the posterior side; widest point of the
metacone on the labial side is shifted anteriorly; posterior portion of the metacone is
compressed; the widest point of the metacone is more anterior; and the width of the
lingual cusps is greater. While these differences are visible in the thin plate spline, they
are also more clear in relationship to the actual tooth in the averaged (consensus)
image (Figure 8).
31
Figure 7. Thin Plate Spline. This shows the differences in morphology when U.
americanus is stretched to fit to the form of U. arctos. The outline of the tooth, labial
cusps, and midline of the tooth have had their landmarks connected for the sake of
recognition.
32
Figure 8. Averaged Images of the M2 of U. americanus (Right) and U. arctos (Left).
These images help to illustrate the visual differences between the teeth, and how
prominent they are between a large number of specimens.
33
The first major difference between the teeth of the two species of Ursus is that
the protocone is more often split in U. americanus. In 27 of 37 (72.97%) specimens of
U. americanus, the protocone is split. It is evident that the protocone is not always split,
and not always split in the same location, by the lack of a clear division, but the cusps
are definitely blurred between landmarks 11 and 12. While some of the same blurring is
seen in U. arctos, it is substantially less. Only 12 of 25 (48%) of U. arctos specimens
had a split protocone. In U. americanus, the labial cusps are more hexagonal, with
points on the lingual side. In U. arctos the paracone is more strictly cuboid and the
metacone is more triangular. While a presence of a post hypocone was recorded to be
more common in U. americanus, both the hypocone and post hypocone are shown to
be highly variable in size and position in both species. This region of the image shows
the general shape of a cusp, but it is only very weakly and any details are very blurry.
While the metaconule is also fairly blurry in both species, it is significantly more robust
and distinct in U. arctos.
Discussion
Morphological Diagnosis
Results from the Principal Components and Discriminant analyses show a strong
indication that the morphology of the M2 of U. arctos and U. americanus can be used to
distinguish the two species. However, an important aspect of this study is creating a
diagnostic technique that doesn’t require a full statistical analysis. While the differences
described in the results can be used as a general guide, the primary focus should be on
the protocone, metacone, and metaconule as these are features with several landmarks
34
specified as important by the PCA and Discriminant analysis. As such, the criteria for
identifying U. americanus and U. arctos by the M2 should go as follows: 1) if the tooth
has a triangular shaped metacone and the metaconule is approximately equal in width
to the metacone, then the tooth should be classified as U. arctos, 2) if the metacone is
cuboid or hexagonal and the metaconule is visibly smaller than the metacone, then the
tooth should be classified as U. americanus. In the case of uncertainty, secondary
characteristics can be used to improve the diagnosis. While these secondary characters
do appear to be indicative of species, they were variable enough that landmarks could
not be regularly placed in locations associated with them. In U. americanus, the
protocone is typically split and both the hypocone and a post-hypocone can usually be
identified. The protocone is only rarely split in U. arctos and a post-hypocone is typically
small to the point of being unidentifiable. With these traits, M2 from either modern or
Pleistocene U. americanus and U. arctos can be classified to at least a conferred
species designation, if not confidently to species.
Further Implications
An unexpected result from the PCA analysis of specimens from both collections
was the separation of the ETVP and NMNH U. arctos specimens. Most of the ETVP
specimens were Kodiak (U. arctos middendorffi), so this may be an indication of a
difference in the morphology of the M2 between the continental brown bear (sampled at
NMNH) and the Kodiak subspecies (ETVP sample). It appears that the U. arctos
middendorffi lingual cusps are more blade like, making their divisions less distinct.
Future studies should be conducted to see if the separation seen in the PCA and visual
35
difference is indicative of a true characteristic, or a trend within the species. The visual
difference might not be consistent enough to be considered as a diagnostic tool, and the
separation in the PCA might be due to a slight difference in the angle the photograph
was taken. If a future study does show a true difference in these M2, then that might
show that the technique is useful amongst a wide variety of Ursus species and
subspecies. Distinguishing U. americanus from U. thibetanus can be difficult at times,
and would likely be further complicated in the fossil record as one approaches their
most recent common ancestor (Larivière 2001). Careful analysis of the M2 may be
useful in distinguishing the two species. In addition, genetic studies show that other
extant species of bear may also be closely related to U. americanus, presenting another
possible candidate for analysis (Kutschera et al. 2014). While this technique has not
been vetted against ursids outside of Ursus, it seems likely that it would be similarly
effective.
Further analysis of the morphology of the M2 could reveal more than just
evolutionary relationships. As the teeth are primarily used for food acquisition and
processing, differences in morphology may be reflective of diet. While U. americanus
and U. arctos are largely omnivorous, there are extant bears adapted for
hypercarnivory, hypocarnivory, insectivory, and herbivory of hard plants (Baryshnikov
2007). Using the characters of the M2 of these different, specialized species could
assist in revealing the diet of extinct bears, such as Arctodus simus and Arctotherium,
whose diet is debated (e.g., Matheus 1995, Sorkin 2006, Figueirido et al. 2010, Meloro
2011). A comprehensive study of the morphology of the teeth of various fossil and
36
extant bear species that gives attention to functional morphology may be useful in
showing why Arctodus and other extinct bears may have different morphologies.
While learning more about the morphology of the M2 could be useful in
distinguishing other species of ursids and learning more about the diets of extinct
ursids, the results of this study can directly be used for future work without secondary
research. With a new method for diagnosing Pleistocene ursines from North America,
specimens from locales which were difficult to diagnose can be reexamined to more
strongly confirm their designation, or correct it. Genetic studies suggest that there may
be specimens which are incorrectly labeled, or there is an important clade of U. arctos
which is relatively absent from the fossil record (Davison et al. 2011). Davison et al.
(2011) found that there should be a clade known as Clade 4 of U. arctos which migrated
south of Beringia into Canada as early as 33,000 cal years BP. Very few fossils of U.
arctos have been reliably identified and dated to older than 12,000 cal years BP
(Matheus 2004). This study could be crucial in identifying any members of this clade
which have been misidentified. A potential similar situation can be seen in the Meachen
et al. (2016) study in which they suggested several wolves from Natural Trap Cave were
misidentified and actually represented the Beringian morphotype. Both the Beringian
wolf case and the Clade 4 of U. arctos, feature a group of animals moving south of
Beringia that have not previously been identified in the fossil record.
37
CHAPTER 3
ECOLOGICAL NICHE MODEL OF U. AMERICANUS AND U. ARCTOS PROJECTED
TO THE LAST GLACIAL MAXIMUM
In addition to the morphological study of the upper second molar, a small
ecological niche study was performed to assist in selecting locations where the results
of said study might be best used. As stated previously, there is potential for an overlap
in U. arctos and U. americanus populations south of the Cordellian and Laurentide ice
sheets as early as 31,000 cal years BP (Davison et al. 2011). The contiguous United
States is quite a range of land in which to search for misidentified fossils, so a method
for narrowing the search radius would be greatly beneficial. This would allow focused
searching for areas where both species are predicted to have occurred, or see where a
species has been identified outside of its expected range. While some areas are known
to not contain U. arctos currently, such as eastern North America (Feldhamer et al.
2009), there are fossil specimens from the Pleistocene that have been identified as U.
arctos as far east as Maine (Graham and Lundelius 2010). As such, it is important to not
disregard that the Pleistocene was a very different time both climatically and
ecologically and modern preconceptions of where these species should not be used.
For an ecological niche study, bioclimatic variables are required. While
bioclimatic variables are readily available from much of recent history, there aren’t any
that span the entirety of the Pleistocene. However, climatic data for the Last Glacial
Maximum (LGM) has been created and is available through WorldClim. The LGM, as
the period with the most glaciation out of the 31,000 years that U. arctos could have
38
been present south of Beringia, acts as a good extreme for comparison with the modern
climate. Something worth noting though, is that while this LGM bioclimatic data is
incredibly useful, it cannot be perfectly accurate. Davis et al. (2014) suggest that
ecological niche model projections of the LGM may have some strong biases. This data
has been generated from a series of paleoenvironmental studies, not collected from
weather stations across the world. As such, it should be subject to scrutiny, and will not
be wholly accurate.
Materials and Methods
Location data for both U. americanus and U. arctos were obtained from the
Global Biodiversity Information Facility (GBIF). Environmental data, in the form of
bioclimatic variables, was obtained from WorldClim, this included both modern data and
data from the LGM. All data processing was performed in Microsoft Office Excel and
ESRI ArcGis. For several specific tasks, the SDMtoolbox add-on for ArcGis, developed
by Jason L. Brown, was used.
Bioclimatic variables were chosen for this study were 1, 4, 12, and 15 which
correspond to Annual Mean Temperature, Temperature Seasonality, Annual
Precipitation, and Precipitation Seasonality, respectively. Only a limited number of the
total possible bioclimatic variables were chosen to avoid the possibility of overfitting.
These particular variables were chosen to represent the general environmental
conditions at the time. Bioclimatic data was downloaded at 5 arc-minutes in ESRI Grid
and CCSM4 types. ESRI Grids were used in the modern model and CCSM4 in the LGM
model. All bioclimatic variables were masked to North America. Species data from GBIF
39
was rarefied to 8 km and masked to North America. Eurasian U. arctos were considered
for inclusion, but ultimately excluded. This was done to avoid any complications caused
by differences in environmental preferences between the North American and Eurasian
clades. While North American clades are descended from Eurasian clades, genetic
studies have shown that there is not any significant level of interbreeding with Eurasian
clades once U. arctos actually extends into North America. Once rarefied and masked,
species data was divided into 80% training and 20% testing subsets.
Both ecological niche models were performed using Maxent. As Maxent is a
relatively easy to use, common, and stable modeling package, it was the software of
choice. Maxent models were ran using 10 percentile training presence as the threshold
rule. Young et al. suggests that this threshold rule is suitable for testing subsets greater
than 10%. Maxent first generated an ecological niche model for the present distribution
of both U. arctos and U. americanus. From the data of the present ecological niche
model, the LGM model could be created using the LGM bioclimatic variables. Rasters
from the four niche models were converted to binary at a .5 threshold. At this threshold,
a presence estimate of less than 50% would be considered absence and an estimate of
greater than 50% would be considered presence. While this does simplify the map, it
allows for comparison between modern and LGM models and helps to counter
overfitting of the model. Presence maps that have not been simplified to binary are still
important, as they show a total range, and are useful in the case that the binary map
under-fits the total range. Once converted to binary, the LGM projection rasters were
reclassified to 1 and 4 as opposed to the 0 and 1 of binary for raster addition and
40
subtractions. Modern model rasters were subtracted from the LGM model rasters to
allow for visual representation of the differences.
Results
The ecological niche models for the modern (Figures 9 and 20) and LGM
(Figures 10 and 11) distributions of both U. americanus and U. arctos can be seen in
the figures below. Accuracy metrics suggest a strong, though imperfect level of
accuracy. Both area under the curve (AUC) values for the Sensitivity versus Specificity
charts (Figure 12) generated by Maxent scored above 8.5, with the Ecological Niche
Model for U. arctos scoring at slightly higher.
The modern ecological niche model for U. americanus (Figure 9) shows the
strongest presence along the west coast, followed by the Rocky Mountains and east
coast with a preference for the upper east coast. The only areas of strong absence are
the northern most portion of Canada and Alaska, the Yucatan Peninsula, and the
Sonoran and Mojave Deserts. For U. arctos, the modern ecological niche model shows
the strongest presence along the Canadian and Alaskan west coast and the Rocky
Mountains. Areas of moderate to light presence include the Chihuahua Mountains in
Mexico and most of Canada reaching to the east coast. Areas of absence include the
Arctic, the middle and eastern areas of the United States, and portions of Mexico and
the western United States.
42
Figure 10. Modern Ecological Niche Model for U. arctos. Note that this model is
restricted to North America, so it does not take the Eurasian range into account.
43
Figure 11. Last Glacial Maximum Ecological Niche Model for U. americanus. This model
represents the range of U. americanus at approximately 21,000 years cal BP.
44
Figure 12. Last Glacial Maximum Ecological Niche Model for U. arctos. This model
represents the range of U. americanus at approximately 21,000 years cal BP.
45
Figure 13. Sensitivity vs. Specificity charts for the Ecological Niche Models. A value of
.5 is equivalent to random chance, with a value of 1 being absolute certainty.
46
The LGM ecological niche model for U. americanus shows strong presence
along the west coast, narrowing to the coast line along the northern portions, the
Chihuahua Mountains, the east of the United States, and portions of most of the
midcontinent. Ursus americanus does show absence across nearly the entirety of
Canada aside from portions of the south and a narrow strip of the west coast. This is
likely due to the presence of the Cordilleran and Laurentide ice sheets spread across
Canada. The LGM ecological niche model for U. arctos shows very strong presence
along the Canadian and Alaskan west coast, most of the United States west coast, and
much of the western half of the United States. There is strong, though lighter presence
throughout much of Mexico, and light presence spreading throughout much of the
continent, all the way to the east coast.
When converted to a binary map, and compared to the modern distribution
(Figures 14, 15, and 16), U. americanus is shown by the ecological niche model to have
a greater range along the west coast and southern range, but reduced presence in the
north. In U. arctos, the ecological niche model shows a much greater area of presence
along the west coast and western half of the United States. This said, there is less
presence along the north-western coast of Alaska and parts of Canada. There are three
principal areas of overlap for the two species in the LGM ecological niche model: A
large portion of the western corner of the United States, a narrow strip along the
Canadian west coast, and a portion of the central United States.
47
Figure 14. Comparison of the U. americanus Modern and LGM Ecological Niche Model.
Areas of high presence (>.5) of U. americanus compared between the current to LGM
model.
48
Figure 15. Comparison of the U. arctos Modern and LGM Ecological Niche Model.
Areas of high presence (>.5) of U. arctos compared between the current to LGM model.
49
Figure 16. Area of significant presence (>.5) overlap between U. americanus and U.
arctos at the LGM (~21,000 cal years BP).
50
Discussion
The ecological niche models have a strong level of accuracy based upon their
AUC scores, and a comparison with the modern historic range of both species shows a
strong similarity. The U. americanus historic range (Pelton et al. 1999) matches the
niche model closely, and while the U. arctos model includes everywhere that U. arctos
occurs, it does seem to over fit as it shows light presence along the east coast
(Feldhamer et al. 2003). This said, the LGM niche model, as well as several fossil
localities suggest that U. arctos used to inhabit this area (Graham and Lundelius 2010).
This suggests that the model may not actually be overfitting, but simply cannot account
for some other environmental factor. A difference in available food or vegetation
between the LGM and present could be the cause of this range reduction. While a
vegetation index can be included in modern ecological niche models, they have not
been created for the LGM. Niche models could be created to account for the presence
or absence of the food items for U. arctos, but as U. arctos can consume a very wide
variety of food items, the scale of that task puts it outside the scope of this study.
The large area of strong presence for U. arctos south of Beringia in the LGM
ecological niche model suggests that U. arctos would have been successful throughout
much of North America. The presence of U. arctos fossils as far east as Maine suggest
that the light presence in that area may have been stronger than indicated, or at least
stronger at a later point in the Pleistocene (Graham and Lundelius 2010).
With a particularly low presence in the southeast of the United States from U.
arctos, and the opposite with U. americanus, this may be the only area south of the ice
sheets where the two species did not coexist, or were unlikely to do so. In all other
51
areas south of the ice sheet, both species have potential presence, but at varying
degrees. As such, identifications should not be made based on location. This is
especially important in the areas of strong overlap, namely the west coast and Rocky
Mountains. These are areas where a careful identification is especially important, as
both species could easily be present. The other two areas of concern are the southeast
and the majority of Canada. These are areas where one species is present, and the
other is largely absent. Specimens found in these areas that have been identified as a
species other than that expected by the ecological niche model should be closely
analyzed based upon the morphological study shown previously. This said, it is possible
that the ecological niche model is not accurately representing these areas, or that the
specimen has wandered from its typical range, so these specimens should not be
identified based solely on their location.
52
CHAPTER 4
CASE STUDY AND CONCLUSIONS
Through a morphometric analysis of the second upper molar, this study has
strived to develop an improved technique for the identification of Ursus species in North
America. In addition, niche models were developed to help identify areas where each
fossil species was likely to occur during the LGM. While the morphological analysis and
the ecological niche models can be used independently, they can also be used in
conjunction to assess the record of LGM bears across North America. The niche
models can be used to identify fossil sites where one or both species of Ursus would be
expected to occur, and the morphological analyses can be used to assist in the
identification of any specimens found at those localities. Additionally, this pair of studies
can be used in conjunction with FAUNMAP to check for localities that report a specimen
outside of its expected range or report one expected species but not the other.
While this study appears successful in generating a technique for identifying
specimens and locating where they are likely to occur, this success is based on AUC
values and degrees of separation. The real degree of success is to be seen in the
successful application of these techniques. FAUNMAP lists six specimens of U. arctos
in the eastern half of North America that date to the Pleistocene. While the ecological
niche model for the LGM shows that there was some presence of U. arctos in this area,
the presence is at a very low number, so fossils from this area might be uncommon.
Two Pleistocene specimens previously identified as Ursus arctos from the
Carnegie Museum of Natural History were borrowed for analysis. This includes CM
12617, from Welsh Cave, KY, and CM 12999 from Organ Cave, WV. Both specimens
53
are reported in the literature as U. arctos (Guilday 1968, Guilday et al. 1977). Guilday
(1968) and Guilday et al. (1977) do not state how these two specimens are identified,
but as CM 12617 is 34mm long and CM 12999 is 33mm long, it seems that if Kurtén
and Anderson’s (1980) method of identification by tooth length was considered, it would
not have been particularly useful. Kurtén and Anderson (1980) stated, and Graham
(1991) restated, that the M2 of fossil U. americanus never exceeded 34mm in length,
and that fossil U. arctos rarely had M2 of such a short length. At 33mm and 34 mm in
length, the M2s of these two specimens would be difficult to identify by tooth length.
Hence, these two specimens were a prime candidate for a case study attempting to
support their identification, or re-identify them.
The method for visually identifying the two species described on page 33, was
applied to both teeth, with mixed results. Images of these teeth can be found in
Appendix A. CM 12617 was determined to be correctly identified based on the presence
of a large metaconule and triangular metacone. CM 12999 was determined to be
incorrectly identified based on a relatively small metaconule, and cuboid metacone. It
should be noted that CM 12999 is unusual in that it appears to have an accessory cusp
sitting posterior to the metacone, potentially compacting it. While this accessory cusp
could be masking a more triangular shape for the metacone, the small size of the
metaconule suggests that this is not the case. As such this study refers CM 12999 to U.
americanus.
In order to strengthen this referral, as this would be the first application of this
technique to a fossil specimen, photographs of CM 12617 and 12999 were landmarked
54
and included in a principal component analysis and discriminant analysis alongside the
specimens from the NMNH and ETVP collections.
The principal component analysis shows very similar eigenvalues and principal
components to the previous principal component analysis (Tables 6 and 7). The two
scatterplots (Figures 17 and 18) show that the fossil specimens plot within the morpho
space of their expected identifications, though they are not nestled closely within that
morpho space.
The discriminant function and stepwise discriminant function show eigenvalues
and Wilk’s lambdas very similar to the previous discriminant functions (Tables 8 and 9).
However, the discriminant function did not misidentify as many specimens as U. arctos,
increasing its success rate somewhat (See Appendix A). Additionally, the fossil
specimens are classified as their expected groups, with CM 12617 classified as U.
arctos and CM 12999 classified as U. americanus (Figures 19 and 20).
The results of the principal component analysis and discriminant functions
indicate that these two fossil specimens do not share the same species morpho space
and are classified differently statistically. This difference in principal component and
discriminant function scores agrees with the results of the morphological diagnosis, so
this study would redesignate CM 12999 as U. americanus. This case study serves of an
example of how this type of study is useful, and exemplifies the difficulty of identification
without it. In the future, this analysis can be used to facilitate the identifications of Ursus
specimens across North America, and potentially uncover previously under represented
clades within U. arctos that have been misidentified.
56
Figure 17. Scatter Plot of the Case Study PC1 and PC2. The first principal component
shows lateral compression or extension with an inverse effect along the anterior
landmarks. The second principal component shows a vertical bending of the anterior
landmarks of the tooth.
57
Figure 4. Scatter Plot of the Case Study PC1 and PC3. The first principal component
shows lateral compression or extension with an inverse effect along the anterior
landmarks. The third principal component shows a twisting such that the lingual and
labial sides of the middle of the tooth inversely compress or expand.
59
Table 8. Eigenvalues (Top), Wilk’s Lambda (Middle), and Classification Results
(Bottom) of the Case. Classification results show a strong level of accuracy.
60
Figure 19. Histogram Plot of the Case Study Discriminant Function. The discriminant
function represents the same variation as the first principal component.
61
Table 9. Stepwise Eigenvalues (Top), Wilk’s Lambda (Middle), and Classification
Results (Bottom) of the Case Study. Classification results show a strong level of
accuracy, though the presence of a possibly misidentified specimen within the collection
lowers the accuracy of U. arctos identification to 96.66%.
62
Figure 20. Histogram Plot of the Case Study Stepwise Discriminant Function. The stepwise discriminant function represents very similar variation as the discriminant function.
63
BIBLIOGRAPHY
Baryshnikov, G. 2006. Morphometrical variability of cheek teeth in cave bears. Scientific
Annals, School of Geology Aristotle University of Thessaloniki (AUTH) 98:79-102.
Baryshnikov G, Germonpré M, Sablin M. 2003. Sexual dimorphism and morphometric
variability of cheek teeth of the cave bear (Ursus spelaeus). Belgian Journal of
Zoology 133(2):111-120.
Baryshnikov, G. 2007. Bears Family (Carnivora, Ursidae). Fauna of Russia and
Neighbouring Countries 147.
Bookstein, F.L. 1991. Morphometric tools for landmark data: geometry and biology.
Cambridge University Press.
Czaplewski NJ, Puckette WL. 2015. Late Pleistocene Remains of an American black
bear (Ursus americanus) and Two Small Vertebrates from an Oklahoma Ozark
Cave 94.
Czaplewski NJ, Willsey S. 2013. Late quaternary brown bear (Ursidae: Ursus cf. arctos)
from a cave in the Huachuca Mountains, Arizona.
Czaplewski NJ, Mead J, Bell CJ, Peachey WD, Ku T. 1999. Papago Springs Cave
revisited, part II: vertebrate paleofauna. Occasional Papers of the Oklahoma
Museum of Natural History.
64
Davis EB, McGuire JL, Orcutt JD. 2014. Ecological niche models of mammalian glacial
refugia show consistent bias. Ecography 37(11):1133-1138.
Davison J, Ho SY, Bray SC, Korsten M, Tammeleht E, Hindrikson M, Østbye K, Østbye
E, Lauritzen S, Austin J. 2011. Late-Quaternary biogeographic scenarios for the
brown bear (Ursus arctos), a wild mammal model species. Quaternary Science
Reviews 30(3):418-430.
DeMaster DP, Stirling I. 1981. Ursus maritimus. Mammalian Species 145:1-7.
Elftman, H.O. 1931. Pleistocene mammals of Fossil Lake, Oregon. American Museum
Novitates 481.
Feldhamer GA, Thompson BC, Chapman JA. 2003. Wild mammals of North America:
biology, management, and conservation. JHU Press.
Figueirido B, Pérez-Claros JA, Torregrosa V, Martín-Serra A, Palmqvist P. 2010.
Demythologizing Arctodus simus, the ‘short-faced’ long-legged and predaceous
bear that never was. Journal of Vertebrate Paleontology 30(1):262-275.
Gordon, K.R. 1977. Molar measurements as a taxonomic tool in Ursus. Journal of
Mammalogy 58(2):247-248
Graham, R.W. 1991. Variability in the size of North American Quaternary black bears
(Ursus americanus) with the description of a fossil black bear from Bill Neff Cave,
Virginia. Illinois State Museum Scientific Papers 23:238-250.
65
Graham R, Lundelius E Jr. 2010. FAUNMAP II: new data for North America with a
temporal extension for the Blancan, Irvingtonian and early Rancholabrean.
FAUNMAP II Database, version 1.
Guilday, J.E. 1968. Grizzly bears from eastern North America. American Midland
Naturalist:247-250.
Guilday JE, Parmalee PW, Hamilton HW. 1977. The Clark's Cave bone deposit and the
late Pleistocene paleoecology of the central Appalachian Mountains of Virginia.
Kurtén B, Anderson E. 1980. Pleistocene mammals of North America. Columbia
University Press.
Kurtén, J. 1963. Fossil bears from Texas. The Pearce-Sellards Series 1:3-14
Kutschera VE, Bidon T, Hailer F, Rodi JL, Fain SR, Janke A. 2014. Bears in a forest of
gene trees: phylogenetic inference is complicated by incomplete lineage sorting and
gene flow. Molecular biology and evolution 31(8):2004-2017.
Larivière, S. 2001. Ursus americanus. Mammalian species 647:1-11.
Matheus, P.E. 1995. Diet and co-ecology of Pleistocene short-faced bears and brown
bears in eastern Beringia. Quaternary Research 44(3):447-453.
Matheus P, Burns J, Weinstock J, Hofreiter M. 2004. Pleistocene brown bears in the
mid-continent of North America. Science 306(5699):1150.
66
Meachen JA, Brannick AL, Fry TJ. 2016. Extinct Beringian wolf morphotype found in the
continental US has implications for wolf migration and evolution. Ecology and
Evolution.
Meloro, C. 2011. Feeding habits of Plio-Pleistocene large carnivores as revealed by the
mandibular geometry. Journal of Vertebrate Paleontology 31(2):428-446.
Mustoe GE, Carlstad CA. 1995. A late Pleistocene brown bear (Ursus arctos) from
northwest Washington. Northwest Science 69(2):106-113
Pasitschniak-Arts, M. 1993. Ursus arctos. Mammalian Species 439:1-10.
Pelton MR, Coley AB, Eason TH, Martinez DD, Pederson JA, van Manen FT, Weaver
KM, SERVHEEN Y. 1999. American black bear conservation action plan.
C.Servheen, S.Herrero, and B.Peyton, compilers.Bears.Status survey and
conservation action plan.IUCN/SSC bear and polar bear specialist groups. 144-156.
Rohlf, F.J. 2015. The tps series of software. Hystrix, the Italian Journal of Mammalogy
26(1):9-12.
Sorkin, B. 2006. Ecomorphology of the giant short-faced bears Agriotherium and
Arctodus. Historical Biology 18(1):1-20.
Taberlet P, Bouvet J. 1994. Mitochondrial DNA polymorphism, phylogeography, and
conservation genetics of the brown bear Ursus arctos in Europe.
Proceedings.Biological sciences 255(1344):195-200.
67
Wolverton S, Lyman RL. 1998. Measuring late Quaternary ursid diminution in the
Midwest. Quaternary Research 49(3):322-329.
68
APPENDIX
Additional Tables and Figures
Specimen Species State or Territory
USNM 236227 U. americanus Colorado
USNM 224509 U. americanus Colorado
USNM 223943 U. americanus Florida
USNM 234242 U. americanus Florida
USNM 227660 U. americanus Idaho
USNM 216420 U. americanus Idaho
USNM A03061 U. americanus New York
USNM 187876 U. americanus New York
USNM 228262 U. americanus New Mexico
USNM 231359 U. americanus New Mexico
USNM 206132 U. americanus Alaska
USNM 087617 U. americanus British Colombia
USNM 081198 U. americanus Quebec
USNM 205950 U. americanus California
USNM 248531 U. americanus Washington
USNM 159368 U. americanus Louisiana
USNM 135141 U. americanus Louisiana
USNM 227926 U. americanus Wyoming
ETVP 18252 U. americanus Tennessee
ETVP 18244 U. americanus Tennessee
ETVP 7170 U. americanus Tennessee
ETVP 7169 U. americanus Quebec
ETVP 5011 U. americanus ?
ETVP 18233 U. americanus Tennessee
NAUQSP 7607 U. americanus Maine
CC.279 U. americanus ?
CC.388 U. americanus ?
NVPL 6918 U. americanus Tennessee
ETVP 7173 U. americanus ?
ETVP 10138 U. americanus Tennessee
ETVP 7179 U. americanus ?
ETVP 10129 U. americanus Tennessee
ETVP 10139 U. americanus Tennessee
ETVP 18175 U. americanus Tennessee
USNM 211240 U. arctos Montana
69
USNM 225621 U. arctos Montana
USNM 098324 U. arctos Chihuahua
USNM 098320 U. arctos Chihuahua
USNM A31276 U. arctos Idaho
USNM 233241 U. arctos Idaho
USNM 113410 U. arctos Colorado
USNM 113411 U. arctos Colorado
USNM 228228 U. arctos Yukon
USNM 227977 U. arctos Yukon
USNM 223945 U. arctos British Colombia
USNM 223689 U. arctos British Colombia
USNM 262374 U. arctos New Mexico
USNM 223034 U. arctos Utah
USNM 228226 U. arctos California
USNM 235445 U. arctos Wyoming
USNM 203524 U. arctos North Dakota
USNM 242652 U. arctos Arizona
USNM 243786 U. arctos Washington
USNM 222107 U. arctos Alberta
ETVP 5162 U. arctos ?
ETVP 18264 U. arctos Alaska
AKGBR 9904282 U. arctos Alaska
AKGBR 0304553 U. arctos Alaska
AKGBR 0502474 U. arctos Alaska
AKGBR 1002343 U. arctos Alaska
AHGBR 0304912 U. arctos Alaska
AKGBR U. arctos Alaska
ETVP 10145 U. arctos ?
ETVP 10501 U. arctos ?
80
VITA
THERON MICHAEL KANTELIS
Education: North Forsyth High School, Cumming, Georgia 2011
B.S. Biology, Berry College, Mount Berry, Georgia 2015
M.S. Geosciences, East Tennessee State University,
Johnson City, Tennessee 2017
Professional Experience: Graduate Assistant, East Tennessee State University,
Department of Geosciences, 2015-2017
Recommended